Introduction

This is a personal Rmarkdown document I have created to visualize the COVID-19 updates and some preliminary exploratory data analysis (EDA). The source of this data is the github repository created and maintained by the Coronavirus COVID-19 Global Cases by the Center for Systems Science and Engineering (CSSE) at Johns Hopkins University (JHU).

suppressPackageStartupMessages(library(tidyverse))
suppressPackageStartupMessages(library(forecast))
suppressPackageStartupMessages(library(zoo))
suppressPackageStartupMessages(library(xts))
suppressPackageStartupMessages(library(gridExtra))
suppressPackageStartupMessages(library(gghighlight))
suppressPackageStartupMessages(library(knitr))
suppressPackageStartupMessages(library(directlabels))
suppressPackageStartupMessages(library(scales))
suppressPackageStartupMessages(library(plotly))
#suppressPackageStartupMessages(library(rjson))

Reading the data

COVID_confirmed_global_raw <- read_csv("csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_confirmed_global.csv") 
COVID_deaths_global_raw <-  read_csv("csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_deaths_global.csv") 
COVID_recovered_global_raw <-  read_csv("csse_covid_19_data/csse_covid_19_time_series/time_series_covid19_recovered_global.csv")

Reshaping and formatting data

# Reshape to longer format
COVID_confirmed_global_longer <- COVID_confirmed_global_raw %>% 
  pivot_longer(cols      = c('1/22/20':names(COVID_confirmed_global_raw)[ncol(COVID_confirmed_global_raw)]),
               names_to  = "date",
               values_to = "n_cases")
  

COVID_deaths_global_longer <- COVID_deaths_global_raw %>% 
  pivot_longer(cols      = c('1/22/20':names(COVID_deaths_global_raw)[ncol(COVID_deaths_global_raw)]),
               names_to  = "date",
               values_to = "n_cases")

COVID_recovered_global_longer <- COVID_recovered_global_raw %>% 
  pivot_longer(cols      = c('1/22/20':names(COVID_recovered_global_raw)[ncol(COVID_recovered_global_raw)]),
               names_to  = "date",
               values_to = "n_cases")


# change column names
colnames(COVID_confirmed_global_longer) <-  c('state', 'country', 'lat', 'long','date', 'n_cases')
colnames(COVID_deaths_global_longer) <-  c('state', 'country', 'lat', 'long','date', 'n_cases')
colnames(COVID_recovered_global_longer) <-  c('state', 'country', 'lat', 'long','date', 'n_cases')

# drop `state` column and create a `new_cases` column
COVID_confirmed_global_longer <- COVID_confirmed_global_longer %>% 
               select(-state)%>% 
               group_by(country, date) %>% 
               summarize(n_cases = sum(n_cases))  
COVID_deaths_global_longer <- COVID_deaths_global_longer %>% 
               select(-state)%>% 
               group_by(country, date) %>% 
               summarize(n_cases = sum(n_cases)) 
COVID_recovered_global_longer <- COVID_recovered_global_longer %>% 
               select(-state) %>% 
               group_by(country, date) %>% 
               summarize(n_cases = sum(n_cases)) 

# convert date columns from character to date format
COVID_confirmed_global_longer$date <- as.Date(COVID_confirmed_global_longer$date, format = '%m/%d/%Y')
COVID_deaths_global_longer$date <- as.Date(COVID_deaths_global_longer$date, format = '%m/%d/%Y')
COVID_recovered_global_longer$date <- as.Date(COVID_recovered_global_longer$date, format = '%m/%d/%Y')

COVID_confirmed_global_longer <- COVID_confirmed_global_longer %>% 
  arrange(country, date) %>% 
  mutate(new_cases = n_cases-lag(n_cases, default = 0))

COVID_deaths_global_longer <- COVID_deaths_global_longer %>% 
  arrange(country, date) %>% 
  mutate(new_cases = n_cases-lag(n_cases, default = 0))

COVID_recovered_global_longer <- COVID_recovered_global_longer %>% 
  arrange(country, date) %>% 
  mutate(new_cases = n_cases-lag(n_cases, default = 0))

Let’s look at the current data format

knitr::kable(head(COVID_confirmed_global_longer),format = 'markdown')
country date n_cases new_cases
Afghanistan 0020-01-22 0 0
Afghanistan 0020-01-23 0 0
Afghanistan 0020-01-24 0 0
Afghanistan 0020-01-25 0 0
Afghanistan 0020-01-26 0 0
Afghanistan 0020-01-27 0 0

creating some functions for quick stats

world_summary <- function() {
   
  df1 <- COVID_confirmed_global_longer %>% 
    group_by(country) %>% 
    summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases)) %>%
    ungroup() %>% 
    summarize(n_cases_total = sum(n_cases_today),
              new_cases_total = sum(new_cases_today))
  
  df2 <- COVID_deaths_global_longer %>% 
    group_by(country) %>% 
    summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases)) %>% 
    ungroup() %>% 
    summarize(n_cases_total = sum(n_cases_today),
              new_cases_total = sum(new_cases_today))
  
  df3 <- COVID_recovered_global_longer %>%
    group_by(country) %>% 
    summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases)) %>% 
    ungroup() %>% 
    summarize(n_cases_total = sum(n_cases_today),
              new_cases_total = sum(new_cases_today))
  
  print(paste0("number of total confirmed cases in the world as of today:  ", df1$n_cases_total, " with ", df1$new_cases_total, " new cases"))
  print(paste0("number of total deaths in the world  as of today:  ", df2$n_cases_total, " with ", df2$new_cases_total, " new deaths"))
  print(paste0("number of total recovered cases in the world  as of today:  ", df3$n_cases_total, " with ", df3$new_cases_total, " new cases"))
  
}


country_summary <- function(country1) {
  
  df1 <- COVID_confirmed_global_longer %>% group_by(country) %>% dplyr::filter(country==country1) %>%  summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases))
  df2 <- COVID_deaths_global_longer %>% group_by(country)%>% dplyr::filter(country==country1) %>% summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases))
  df3 <- COVID_recovered_global_longer %>% group_by(country)%>% dplyr::filter(country==country1) %>% summarize(n_cases_today = max(n_cases),
              new_cases_today = dplyr::last(new_cases))
  #                                                            
  print(paste0("number of confirmed cases in ", country1, " as of today:  ", df1$n_cases_today, " with ", df1$new_cases_today, " new cases"))
  # df1$n_cases_today

  print(paste0("number of deaths in ", country1, " as of today:  ", df2$n_cases_today, " with ", df2$new_cases_today, " new deaths"))
  # df2$n_cases_today

  print(paste0("number of recovered cases in ", country1, " as of today:  ", df3$n_cases_today, " with ", df3$new_cases_today, " new cases"))
  # df3$n_cases_today
  
}


world_summary()
## [1] "number of total confirmed cases in the world as of today:  1095917 with 82760 new cases"
## [1] "number of total deaths in the world  as of today:  58788 with 5804 new deaths"
## [1] "number of total recovered cases in the world  as of today:  225846 with 15533 new cases"
country_summary("US")
## [1] "number of confirmed cases in US as of today:  275586 with 32133 new cases"
## [1] "number of deaths in US as of today:  7087 with 1161 new deaths"
## [1] "number of recovered cases in US as of today:  9707 with 706 new cases"
country_summary("Italy")
## [1] "number of confirmed cases in Italy as of today:  119827 with 4585 new cases"
## [1] "number of deaths in Italy as of today:  14681 with 766 new deaths"
## [1] "number of recovered cases in Italy as of today:  19758 with 1480 new cases"
country_summary("Spain")
## [1] "number of confirmed cases in Spain as of today:  119199 with 7134 new cases"
## [1] "number of deaths in Spain as of today:  11198 with 850 new deaths"
## [1] "number of recovered cases in Spain as of today:  30513 with 3770 new cases"
country_summary("China")
## [1] "number of confirmed cases in China as of today:  82511 with 79 new cases"
## [1] "number of deaths in China as of today:  3326 with 4 new deaths"
## [1] "number of recovered cases in China as of today:  76760 with 195 new cases"
country_summary("Egypt")
## [1] "number of confirmed cases in Egypt as of today:  985 with 120 new cases"
## [1] "number of deaths in Egypt as of today:  66 with 8 new deaths"
## [1] "number of recovered cases in Egypt as of today:  216 with 15 new cases"

Donut plot

df <- COVID_confirmed_global_longer %>% mutate(country_sum = ifelse(n_cases > 5000, country,"other"))
df <- df %>% group_by(country_sum)
df <- df %>% summarize(count = max(n_cases))
fig <- df %>% plot_ly(labels = ~country_sum, values = ~count, text = ~country_sum)
fig <- fig %>% add_pie(hole = 0.4)
fig <- fig %>% layout(title = "Confirmed cases worldwide",  showlegend = F,
                      xaxis = list(showgrid = FALSE, zeroline = FALSE, showticklabels = FALSE),
                      yaxis = list(showgrid = FALSE, zeroline = FALSE, showticklabels = FALSE))

fig

Multi-dimensional bar-chart

COVID_confirmed_global_longer %>%
  group_by(country) %>%
  plot_ly(x = ~date, y = ~n_cases, color = ~country) %>%
  add_bars(text = ~country)%>%
  layout(barmode = "stack",
         showlegend = FALSE) 
## Warning in RColorBrewer::brewer.pal(N, "Set2"): n too large, allowed maximum for palette Set2 is 8
## Returning the palette you asked for with that many colors

## Warning in RColorBrewer::brewer.pal(N, "Set2"): n too large, allowed maximum for palette Set2 is 8
## Returning the palette you asked for with that many colors

A function for plotting COVID-19 cases for different countries

plot_countries <- function(df, curve_title, cumulative=TRUE, ...) {
  df1 <- df %>% 
    dplyr::filter(country %in% list(...))
  
  if (cumulative) {
  p1 = ggplot(df1, aes(date, n_cases, group=country, color=country))+
    geom_line()+
    scale_x_date(date_breaks = "3 days")+
    scale_y_log10(labels = function(x) format(x, scientific = FALSE),
                  name = "number of cases",
                  breaks = scales::breaks_log(n = 10))+
    theme_bw()+
    theme(axis.text.x = element_text(angle = 90), legend.position = "none")+
    ggtitle(curve_title)+
      geom_dl(data = df1, aes(label = country), method = list(dl.combine("first.points", "last.points"), cex = 0.8))
  } else{
  p1 = ggplot(df1, aes(date, new_cases, group=country, color=country))+
    geom_line()+
    scale_x_date(date_breaks = "3 days")+
    scale_y_log10(labels = function(x) format(x, scientific = FALSE),
                  name = "number of cases",
                  breaks = scales::breaks_log(n = 10))+
    theme_bw()+
    theme(axis.text.x = element_text(angle = 90), legend.position = "none")+
    ggtitle(curve_title)+
      geom_dl(data = df1, aes(label = country), method = list(dl.combine("first.points", "last.points"), cex = 0.8))
  }
  

    
  return(p1)
}

plot_countries(COVID_confirmed_global_longer, curve_title = "Confirmed cases (cumulative)", cumulative = TRUE, "US", "Italy", "Canada", "Egypt",  "china")

plot_countries(COVID_deaths_global_longer, curve_title = "Death cases (cumulative)", cumulative = TRUE,"US", "Italy", "Canada", "Egypt",  "china")

plot_countries(COVID_recovered_global_longer, curve_title = "Recovered cases (cumulative)",cumulative = TRUE, "china","US", "Italy", "Canada", "Egypt",  "china")

plot_countries(COVID_confirmed_global_longer, curve_title = "New confirmed cases", cumulative = FALSE,"US", "Italy", "Canada", "Egypt",  "china")

plot_countries(COVID_deaths_global_longer, curve_title = "New death cases", cumulative = FALSE,"US", "Italy", "Canada", "Egypt",  "china")

plot_countries(COVID_recovered_global_longer, curve_title = "New recovered cases", cumulative = FALSE,"US", "Italy", "Canada", "Egypt",  "china")

Visualizing confirmed cases with new cases

Inspired by this minuteearth video. The thing about this visualization is that it doesn’t plot the Cumulative number of confirmed cases with time, instead with the number of new cases on a log-scale, which is more intuitive. Multiple comparisons between countries with very different number of cases could be very made very clear, and it is very easy to detect whether things are getting better.

COVID_confirmed_smoothed <- COVID_confirmed_global_longer %>% 
  tidyr::nest(-country) %>% 
  dplyr::mutate(m = purrr::map(data, loess,
                                               formula = new_cases ~ n_cases, span = 0.5),
                fitted = purrr::map(m, `[[`, "fitted"))

COVID_confirmed_smoothed <- COVID_confirmed_smoothed %>%
  dplyr::select(-m) %>%
  tidyr::unnest()

COVID_confirmed_smoothed2 <- COVID_confirmed_smoothed %>% 
  dplyr::filter(country %in% c("US", "China", "Italy", "Korea, South", "Iran", "Egypt"))

ggplot(data = COVID_confirmed_smoothed2, aes(n_cases, fitted))+
  geom_path(data = COVID_confirmed_smoothed2,aes(n_cases,fitted,color = country, group = country))+
  theme_bw()+
  ylab("number of cases")+
  scale_y_log10(labels = function(x) format(x, scientific = FALSE))+
  scale_x_log10(labels = function(x) format(x, scientific = FALSE))+
  geom_dl(data = COVID_confirmed_smoothed2, aes(label = country), method = list(dl.combine("first.points", "last.points"), cex = 0.8))+
  xlab(label = "Total confirmed cases")+
  ylab(label = "number of new cases")+
  theme(legend.position="none")